New Tone - AI Facility Operations Copilot

Company: New Tone - AI Facility Operations Copilot

Self-reported basis: The description is entirely self-reported by the author, unverified, and lacks any evidence of revenue, customers, or operational traction.

What it appears to be: A mobile-first AI copilot for facility teams that converts technician reports (text, voice, images) into structured work orders using GPT-5.6 with human review, including duplicate detection, recurrence analysis, and multilingual reporting.

Key change: The author describes a shift from isolated maintenance tickets to an AI-enhanced system with bounded historical memory for each asset.

Single most important open question: Does the described AI workflow actually function reliably in real-world facility environments, or is it a prototype that has not yet been tested at scale?

• The description states that New Tone is an AI Facility Operations Copilot.
• It supports technician reporting through:
  • Text input
  • Voice input (speech-to-text)
  • Optional photo attachment
• It uses GPT-5.6 with structured outputs to draft work orders, including:
  • Title
  • Category and priority
  • Summary
  • Observations
  • Potential risks
  • Clarifying questions
  • Recommended checks
  • Duplicate detection
  • Recurrence analysis
  • Evidence classification
• The AI draft is never auto-applied; it requires explicit technician confirmation.
• It integrates with a mobile app built in Flutter/Dart, and an AI layer using a Go gateway that calls OpenAI’s Responses API.
• The system includes:
  • Facility Memory (bounded history of 30 days for each asset)
  • Evidence-to-action separation (e.g., “reported now”, “visible in photo”, “AI inference”)
• It supports PDF protocol generation, assignment, closure, and signature.

Inference: The product is described as a structured AI-assisted work-order creation tool embedded into an existing operational workflow. It is not a standalone chatbot or assistant but a system that integrates with field reporting.

• The author states the inspiration behind New Tone: facility teams treat every maintenance report in isolation, leading to missed recurring issues.
• The product is positioned as a way to give facility teams a safer and more structured way to turn field reports into actionable maintenance work.
• It claims to enable:
  • A technician seeing one incident
  • New Tone seeing the history of the equipment
• The author describes the system as:
  • Not autonomous (human review required)
  • Not a chatbot (structured workflow)
  • Focused on contextualizing maintenance reports with historical data and AI inference
• The long-term goal is to give every piece of equipment a memory, implying a vision for broader facility intelligence.

Inference: The positioning evolved from solving the problem of isolated reporting to building a system that contextualizes work orders using bounded historical memory and AI-assisted reasoning. It is not a general-purpose AI assistant but a niche tool for facility operations.

• The description states that New Tone is built for facility teams.
• These teams manage:
  • Buildings
  • Floors
  • Rooms
  • Equipment
  • Technicians
  • Maintenance work orders
• The primary user is a technician, who reports incidents and creates work orders.
• The system supports:
  • Asset-level reporting
  • Work order lifecycle (NEW → IN_PROGRESS → CLOSED)
  • Multilingual reporting
• It is not described as targeting facility managers or executives directly, but rather the field-level users who create and act on maintenance tickets.

Inference: The ICP appears to be field technicians in facility operations, with a focus on environments where equipment maintenance is frequent and historical context is valuable. No evidence of customer segmentation beyond this.

• Not evidenced.
• There is no mention of pricing, licensing, subscriptions, or monetization strategy.
• The description does not indicate whether the tool is intended for internal use, SaaS, or a one-time deployment.
• The system uses OpenAI API keys and GPT-5.6, but no information on cost structure or usage limits.

• Built with:
  • Flutter (mobile app)
  • Dart
  • Go gateway
  • OpenAI Responses API (GPT-5.6)
  • Structured Outputs
  • REST API
  • Computer vision
  • Speech-to-text
  • PDF generation
• The AI layer:
  • Uses a stateless Go gateway
  • Calls OpenAI with strict JSON schema constraints
  • Validates inputs and outputs
  • Does not persist data (store: false)
  • Uses human-in-the-loop design
• The system supports:
  • Voice input
  • Photo attachments
  • Manual fallback
  • Editable AI drafts
• The architecture is described as:
  • Human-directed
  • With manual review and confirmation required
  • No auto-execution

Inference: The technical stack suggests a lightweight, mobile-first system with AI integration, built for reliability and safety. It is not a complex SaaS platform but a focused tool with clear boundaries.

• Not evidenced.
• There is no mention of:
  • Customers
  • Revenue
  • Adoption
  • Usage metrics
  • Product maturity beyond prototype stage
• The project is described as a working prototype and demonstration.
• The author mentions:
  • Build Week completion
  • Git baseline preservation
  • Real Android-device testing
  • Automated tests
  • Production deployment checks

Inference: The system is at the prototype or demo stage, not yet proven in production environments. No evidence of traction or real-world usage.

• Not evidenced.
• There is no mention of:
  • Competitors
  • Market positioning relative to others
  • Existing tools in facility management or maintenance work-order systems
• The author does not reference any market analysis, competitive landscape, or differentiation strategy.

• Prototype vs. Production: The system is described as a prototype; no evidence of real-world testing or scalability.
• AI Reliability: While structured outputs are used, the system still relies on GPT-5.6, which may introduce hallucinations or inconsistencies.
• Human-in-the-loop Design: The system requires manual confirmation at every step, which may slow adoption or reduce AI utility.
• Data Privacy & Security: OpenAI API is used with no mention of data handling policies, encryption, or compliance.
• Limited Scope: The system is focused on a narrow use case (work order creation) and does not appear to integrate with broader facility platforms.

1. What are the actual performance metrics for GPT-5.6 in this specific workflow? How often do AI-generated drafts require manual correction?
2. Has the system been tested with real technicians or only in prototype form?
3. Are there any known edge cases where AI inference fails or misclassifies evidence?
4. What is the plan for integrating with existing facility management systems (e.g., CMMS)?
5. How does the system handle data governance, especially around sensitive facility information and technician reports?
6. What are the long-term plans for monetization or scaling beyond a single developer team?